Among the transfers from the federal government to the municipalities, there are several sources of funds: constitutional transfers, legal transfers, transfers from the SUS and voluntary transfers through agreements. The work by Soares and Melo (2016) presents a detailed review of the concepts and purposes of each of these resource sources. However, this study is limited to voluntary transfers.

Pursuant to Decree No. 6,170 (2007), "voluntary transfers" represent transfers of treasury resources by the Union to the States, Federal District, and Municipalities, through the execution of agreements and/or transfer contracts, agreements, adjustments or other instruments similar. “Voluntary transfers” were intended to finance the performance of works and/or services of common interest within the three spheres of government (Brasil, 2016).

Ferreira and Bugarin (2007) analyzed intergovernmental transfers in 2090 municipalities in the period from 1995 to 2004, using the panel data method with fixed effects, and confirming the importance that voters attributed to the performance of the economy at the time of voting. The results also suggest that intergovernmental transfers in Brazil were influenced by party politics. This logic seems to suggest and guide the conduct of politicians who, in the exercise of power, try to demonstrate strength and economic development in times close to elections (Ferreira & Bugarin, 2007).

Veiga and Veiga (2013) analyzed increases in intergovernmental transfers and changes in votes associated with the Portuguese legislative elections, where the dataset used covered 278 municipalities in mainland Portugal, with annual data from 1979 to 2005. the Generalized Moments Method (GMM) for a panel of fixed effects specific to the municipality and the election, and thus, the results found suggest that the increase in transfers from the central government to municipalities, in an election year, guarantees additional votes, and that these transfers are targeted at jurisdictions where the government is at risk of losing support. Also in this direction, Bijos (2013) and Moutinho and Kniess (2017) analyzed the causes of voluntary transfers and found evidence that political-electoral motivation, although a significant element in the distributive process, is not justified as the only factor. Thus, Moutinho and Kniess (2017) concluded that the distribution of resources to Brazilian municipalities and states works as an announced rescue mechanism for those locations that need financial assistance and with preponderant political pressure power.

The term 'flypaper effect' emerged from the empirical work of Henderson (1968) and Gramlich (1969) in the United States, where they established demand equations, based on the utility of the representative citizen, subject to the restriction of their “total income”, or that is, the sum of private income and the portion corresponding to intergovernmental transfers without counterpart (Almeida, 2016).

Rios and Costa (2005) confirm the existence of the flypaper effect in 308 Portuguese municipalities. Subsidy money tends to go to whoever joins first, that is, it is willing to increase public capital expenditures compared to current expenditures (Rios and Costa, 2005). Mattos, Rocha, & Arvate (2011) analyzed a reinterpretation of the flypaper effect in Brazilian municipalities and found evidence that higher transfers from the federal government may induce less efficiency in local tax collection compared to increases in private income. Unconditional resource grants do not make municipalities more efficient, they also found evidence that municipalities also seek to increase their additional revenues through their own resources, including the implementation of new taxes, but in general, municipalities that receive more resources are less efficient (Mattos et al., 2011).

In Canada, Mehiriz and Marceau (2014) exposed those grants have a flypaper effect and asymmetric results for the municipalities of Quebec. In Brazil, Almeida (2016) notes the incidence of the flypaper effect in intergovernmental transfers related to Brazilian municipalities, between 2002 and 2010. Also in Brazil, Paz et al. (2021) analyzed data from 5,570 municipalities in the period from 2005 to 2018, and in the analysis of the flypaper effect, verified data such as tax revenue, current expenditure, current transfer, municipal population and GDP, and concluded that the economic and fiscal reality of the municipalities analyzed does not align with the precepts of the flypaper effect (Paz et al., 2021).

Regarding the flypaper effect, Bastida et al. (2021) developed a survey in Central America, which involved data from 298 Honduran municipalities from 2002–2015. Bastida et al. (2021) found evidence that their partisanship is considered the strongest explanatory factor for the allocation of resources, the electoral cycle did not prove to be significant, party alignment also had a weak and positive impact on transfers, but found a strong effect flypaper in Honduras (Bastida et al., 2021). In the study by Rios et al. (2021), evidence of the flypaper effect was also found in Turkey. Given the above, this study considered that expenses were classified separately as current expenses and capital expenses, and thus the first hypothesis is presented.

H1 – The increase in voluntary transfers generates the flypaper effect on expenses.

In the national literature, on the flypaper effect, studies by Sakurai (2013), Moutinho (2016), Soares and Melo (2016), Moutinho and Kniess (2017) and Paz et al. (2021).

Sakurai (2013) analyzed whether data from Brazilian municipalities in the period from 1989 to 2005 on variations in government transfers symmetrically affected the performance of municipal expenditures executed by the regional public administration. Thus, it sought to determine whether these expenditures were susceptible to the flypaper effect, and suggested, with the results obtained, that not only did government transfers had an asymmetric impact on regional public expenditures, but that, above all, this effect would generate a recomposition between current accounts and investments, in addition to being subject to the interference of political factors.

Moutinho (2016), in turn, found empirical evidence by pointing out partisan political determinants in the implementation of voluntary transfers, shown in the work structuring the classification of voluntary transfers from the Union to Brazilian municipalities, in 2012, via individual parliamentary amendments. For this, Moutinho (2016) used the method, or strategy, of documental research, which at the end of the analyzes, indicates that there are advantages gained by the influences of parties that are in power. And so, in municipalities where there is an electoral predominance of opposition parties, there is a lower volume of transfers (Moutinho, 2016),

Soares and Melo (2016) analyzed voluntary transfers from the Union via agreements on party-political, redistributive and technical criteria for Brazilian municipalities in the period between 1995 and 2010. The conclusions of this research indicated that the partisan position of mayors in relation to the Presidency of the Republic can decisively influence the number of transfers received, that is, voluntary transfers from the Union to municipalities are not redistributive, but associated with political and technical factors (Soares & Melo, 2016).

In this sense, Moutinho and Kniess (2017) also found that there is a strong correlation between the dimension of unencumbered resources and the confluence of voters (voters), between the dimension of unencumbered resources and the party cradle of local managers (mayors), who are struggling to obtain these resources. The investigation used official data from the Federal Government, and also prioritized the documental research method in order to analyze voluntary transfers, and other transfers that occurred through Parliamentary Amendment, to Brazilian municipalities, referring to the proposals registered in the year 2012 Regarding partisanship and the flypaper effect, Bastida et al. (2021) found evidence that, in Central America, partisanship has a positive relationship with resource allocation.

In view of the empirical results on political coalitions, the second research hypothesis is presented.

H2 - Political coalition between the party of the president of the republic and the party of the mayor increases the flypaper effect in Brazilian municipalities.

Araujo and Siqueira (2015) evaluated the demand function for regional public goods and services, which included factors considering fiscal illusion based on the median voter theory, further supported by data concerning the year 2010 drawn from a sample of 5,249 municipalities. The authors also proposed to investigate the effects of fiscal illusion on the demand for municipal public spending in Brazil, and concluded that fiscal illusion, correlated to the tax system and fiscal subsidies, can cooperate with the understanding of the increase in public spending, defined by public expenditure, which would be the composite of expenditures made by public entities to subsidize public services (current expenditure) serving society or for the execution of investments (capital expenditure) (Araujo & Siqueira, 2015). In view of the above, the third and fourth hypotheses of this research are presented.

H3 - The lower the tax complexity, the higher the expense.

H4 - The lower the fiscal visibility, the higher the expense.

Following Araújo and Siqueira (2015) and Rios and Costa (2005), the variables complexity and fiscal visibility, referred to in the hypotheses, serve to emphasize the impact of the flypaper effect. For most of the literature, the indicator chosen to measure the degree of complexity and fiscal visibility of tax revenue is the Hirschman-Herfindahl index (HHI), as it is an accurate measure of revenue concentration, as seen since the studies by Tuckman and Chang. (1991) and by Araújo and Siqueira (2015).

Regarding the tax budget in Brazil, Menezes & Pederiva (2015) concluded that just making the budget tax (IO) will not solve Brazilian budget problems, since the analysis of the proposals for IO shows that, apparently, they aim to meet only the immediate interests of parliamentarians, by making the execution of their individual amendments mandatory. The authors argue that the implementation of the tax budget should still require changes in the relationship between the Executive and Legislative powers, because, in order to obtain the expected results, it would be important that the political system in which it will be inserted be modified.

In this scenario, considering that CA 86 and 95 were inserted into the Brazilian Constitution to improve the process of parliamentary amendments to the budget, which, in practice, made parliamentary amendments mandatory by limiting their discretion in handling the budget, in a way that is expected that, in the period after the aforementioned CA, there should be a reduction in the flypaper effect. In view of the above, the fifth hypothesis is presented.

H5 - After Constitutional Amendment 86/2015, there was a reduction in the flypaper effect for municipal expenses.

The five hypotheses tested in this study were formulated based on the flypaper effect theory, discussed and tested in research by Cossio and Carvalho (2001), Rios and Costa (2005), Inman (2008), Mattos et al. (2011), Parmagnani (2011), Sakurai (2013), Mehiriz and Marceau (2014), Araujo and Siqueira (2015), Almeida (2016), Ryu (2017), Paz et al. (2021), Bastida et al. (2021) and Rios et al. (2021). And all five hypotheses tested in this study take into account expenditure, with current expenditure and capital expenditure seen separately.

In order to verify the existence of the flypaper effect in voluntary transfers from the union to Brazilian municipalities, under the context of factors such as political coalitions, fiscal complexity, fiscal visibility and Constitutional Amendment 86, amended by CA 95/2019, data was collected from 5,484 Brazilian municipalities for the periods from 2008 to 2016, in a total of 49,356 initial observations. After eliminating missing values ​​referring to the variables used in the study, a final sample was obtained with 26,619 observations referring to the municipalities, which was used to test the five research hypotheses.

Specifically, financial data was obtained from the publications of Finanças Brasil – Accounting Data for the Municipalities (FINBRA), a database linked to the National Treasury Secretariat (STN) and which contains detailed information on budget execution (income and expenses) and balance sheet for the municipalities. The political data referring to the parties and coalitions of the elected mayors were extracted from the Superior Electoral Court (TSE) and refer to the municipal executive elections that took place in 2004, 2008 and 2012. Finally, the data from the annual population estimates were taken from the Brazilian Institute of Geography and Statistics (IBGE).

All financial variables were corrected by the national consumer price index (IPCA), using the year 2008 as the base year. This procedure allows the temporal comparability of the financial information as all the variables started to be worked at the same price level (2008 current prices). The financial data were also winsorized at 1% in each tail in order to mitigate the possible influence of outliers and in order to guarantee comparability with other results in the literature.

The analyzes were performed with panel data to test the flypaper effect on current and capital expenditure accounts of Brazilian municipalities, according to models 1 and 2, respectively. The unbalanced panel data models relate the Brazilian municipalities i = 1.2, 3..., 5,484 in the time period t = 2008, 2009, ..., 2016, namely:

(1)

(2)

In order to verify whether the most appropriate panel estimator to estimate models 1 and 2 should be the fixed effects (EF) or random effects (EA) estimator, the Hausman test was used separately for each one of the specifications. The Hausman test consists, in this case, of estimating at the first moment using both estimators (EA and EA). In a second moment, through a statistic with chi-square distribution, it is then verified whether the joint difference between the coefficients of both estimated models is significant. If the null hypothesis of unsystematic differences between the coefficients is rejected, the EF estimator should be used instead of the EA estimator (EF is consistent and EA is inconsistent). On the other hand, if the null hypothesis of unsystematic differences between the coefficients is not rejected, the EA estimator should be used instead of the EF estimator (EA is consistent and more efficient) (Wooldridge, 2002).

The results obtained from the estimation of the Hausman test in each model indicate that the regression of model 1 must be estimated using the fixed (consistent) effect estimator, while the random effects estimator must be used to estimate the regression of the model 2 (consistent and more efficient estimates). In both estimations, robust estimators were used for variance (Wooldridge, 2002; Hsião, 2014).

According to hypothesis H1, the increase in voluntary transfers is expected to generate the flypaper effect on expenditures, whether capital or current. In this case, it is expected that the coefficient β_1 of models 1 and 2 are both positive. In these models, the variables represent changes in capital and current expenditures, in relation to changes in voluntary current transfers for each municipality, and the variable represents the variation of GDP per capita of the municipalities in relation to the variation of voluntary capital transfers.

The coefficients, in turn, respond to hypotheses H2, H3, H4 and H5, which refer to the sensitivity of the flypaper effect to the bloc variables between the party of the president of the republic and the party of the mayor, fiscal complexity, fiscal visibility and the institution of Constitutional Amendment 86 of 2015. According to these hypotheses, and are expected.

Chart 1: Description of the variables used

Based on the study by Inman (2008), the population of the municipality was included as a control variable, as the size of the municipality can influence public spending, which delimits its levels. Another point to be considered is the volume of resources needed to maintain the municipality with its structure in operation, as population growth may vary from municipality to municipality and over time.

Table 01 presents in a detailed way the descriptive statistics of the variables used in this work by year in the period from 2008 to 2016. Analyzing the independent variables referring to income and visibility of taxes, it appears that the average per capita income and visibility of direct taxes showed an upward trend in the period considered. Specifically, the average per capita income over the years presented an average growth of close to 4.5%, although it showed a drop in 2013. Current expenses follow a similar movement to population growth. Such movements initially suggest a certain stability in the growth of current expenses in relation to the increase in population demand, this stability can be explained by the pressures existing in the period due to the political and economic turmoil experienced by Brazil, in which the control of expenses became a constant agenda of analysis and annotation by the media, specialists, and the population. Despite this, the international literature (Lupu & Asandului, 2017; Trabelsi, 2018) and national literature (Bogoni et al., 2011; Degenhart et al., 2016) found evidence of a positive relationship between economic growth and public spending.

Regarding capital expenditures, there was a high average peak in 2010, with a sharp drop in 2011 and subsequent growth. This movement is similar to the variations in average capital transfers and is in line with attempts to formulate investment policies in infrastructure as a way to expand economic activity and contain the effects of declines in economic activity levels (Lupu & Asandului, 2017; Trabelsi, 2018; Bogoni et al., 2011; Degenhart et al., 2016).

Table 01: Descriptive Data Statistics

Source: Research data. D.V refers to the standard deviation. Max refers to the maximum value. Min refers to the Minimum value. Data were winsorized at 1% in each tail.

Both the maintenance of growth in government spending, and the maintenance of mandatory expenditure, limits the government's ability to alleviate taxation on companies and other taxpayers. In this case, although some proposals have emerged to simplify the tax model, there is still high taxation based on a tax system of relatively high complexity. These points are evidenced by the evolution of tax complexity and fiscal visibility, which had a similar behavior in terms of average growth over the years analyzed, with a high peak in 2013, a considerable drop in 2014, and growth in consecutive years.

Although such characteristics are observed in average terms, there is a great oscillation between the analyzed municipalities, evidenced not only by a matter of dispersion, but by the asymmetry present in the analyzed distributions. This portrait is common in the Brazilian scenario, since there are municipalities that demand a greater volume of resources, in addition to having a high population density and greater concentration of wealth, as is the case of the cities of São Paulo and Rio de Janeiro, among others (Instituto Brazilian Institute of Geography and Statistics - IBGE, 2018).

With such a discrepancy between the data, Barnett and Lewis (1994) recommend that the winsorization of the information used be carried out, which consists of replacing the data below and above a given percentage, respectively, with the highest and lowest remaining value. In this study, winsorization was performed at 1%.

Non-tabulated results show that the variables CF and FV have a positive Pearson correlation of 0.85, which indicates that fiscal complexity and visibility have moved very similarly over time. Although these variables showed a relatively high correlation, non-tabulated results for the Variance Inflation Factor (VIF) test showed a value of less than three for the VIF of the first regression and close to four for the second, thus not showing multicollinear implications for any of the regressions. regressions (Gujarati, Porter, & Gunasekar, 2011).

Results not tabulated on Pearson's Correlation also show an inverse relationship between per capita income and the level of current and capital transfers. Such results are in line with the constitution insofar as they reflect the assistential role of voluntary transfers, as recommended by legislation through the delivery of current or capital income to another entity of the Federation, as cooperation, aid or financial assistance (Law n. 101, 2000). In November 2016, the Voluntary Transfers Forum organized by the Ministry of Planning, Development and Management noted that: (i) there are more than half a million proposals for agreements in the Agreement System (Siconv); (ii) in the last 15 years, the Union transferred more than BRL 80 billion to municipalities (Brasil, 2016).

In Table 2 it is possible to visualize the results of the regression estimations that consider capital and current expenditures as dependent variables, respectively, as well as diagnostic evaluation tests of the regressions assumptions. The estimated models did not present multicollinearity problems as the mean VIF was less than 5 for all models (Wooldridge, 2002).

It was also found that in such estimates the data are heteroscedastic, so a robust estimator was used for the variance in both estimates. The Hausman test was used to verify whether the panel to be estimated should contain fixed or random effects. The results suggested that model 1 estimated with fixed effect is consistent, while consistent and more efficient results are obtained with random effects for regression 2, which considers current expenditures as a dependent variable.

Table 2: Results of the estimations

Source: Research data. Income/TVUC income and voluntary transfers from agreements, CF refers to fiscal complexity. VF refers to fiscal visibility. Models with fixed effects most indicated in regression 1 and random effects for regression 2. The term constant refers to the mean of the intercepts of the n municipalities for model 1. The symbol *** denotes significant at 1%; ** denotes significant at 5%; * denotes significant at 10%.

The results shown in Table 2 indicate that the income/TVUC variable is significant in both estimates, although it shows opposite signs. Specifically, changes in income in relation to voluntary transfers have a negative association with current expenditures and a positive association with capital expenditures. Such analyzes suggest that the increase in income is associated with a reduction in the current expenditure of municipalities and an increase in capital expenditure. Therefore, there is a flypaper effect for capital expenditures and not for current expenditures. In this case, there is evidence that supports hypothesis H1 of the presence of the flypaper effect within the perspective of capital expenditures, but not current ones, which is in line with the research by Araujo and Siqueira (2015) and Rios and Costa (2005).

Rios and Costa (2005) confirm the essence of the flypaper effect (that the money from subsidies tends to adhere to where it arrives first, that is, it is willing to increase public expenditure, as opposed to the values passed on to voters) in the analysis of intergovernmental transfers to 308 Portuguese municipalities. In their analyses, the authors conclude that this effect is more prevalent in capital expenditures than in current expenditures, which is in line with the results obtained for Brazilian municipalities. In this case, it appears that the flypaper effect in Brazil is a fact arising from the fixation of resources originating from lumpsum-type government transfers, which are applied to complement local public expenditure instead of promoting the return of resources to the population, for example, through tax relief (Inman, 2008).

The results of the estimation of model 1 and shown in Table 2 also indicate that in municipalities where the ruler is from the same party or coalition as the president, the transfers in capital expenditure are, on average, higher by R$ 566.16 in relation to municipalities that do not show such a relationship. This result is in line with hypothesis H2, in which the coalition between the president's party and the mayor's party increases the volume of current and capital expenditures, concomitantly with the increase in voluntary transfers from agreements. The results presented in Table 2 corroborate the findings of studies by Soares and Melo (2016), Moutinho and Kniess (2017) and Bastida et al. (2021) regarding the effect of party-political criteria on voluntary transfers.

Together, the results on the income effect and the government alignment effect suggest that voluntary transfers from the Union may legally reflect political maneuvers with the legal subterfuge of financing the performance of works and/or services of common interest within the scope of the three spheres of government. (Brazil, 2011). This may even be one of the justifications for the initial evidence found in the literature that the increase in unconditional lump-sum transfers proportionally encourages more municipal public expenditure than the increase in income (Rios & Costa, 2005).

The results in Table 2 also indicate that greater fiscal complexity tends to reduce capital expenditures in the municipalities studied, as can be seen in the estimates of model 1. Specifically, an increase of 0.1 percentage point would tend to reduce capital expenditures by 1.28 percent of total transfers received, which supports hypothesis H3, in which lower fiscal complexity results in increased municipal capital expenditures. The same results are not obtained for current expenses, which remain unrelated to fiscal complexity. This result suggests that changes in tax complexity may favor investment in structural goods, since this demand has even been the target of legislation proposals with the aim of simplifying the tax system and reducing the negative externalities generated by different state and municipal legislation. The results about the effect of fiscal visibility on the variation of capital and current expenditures follow in an analogous way and support the evidence of fiscal complexity. Hypotheses H3 and H4, therefore, are still satisfied for capital expenditures, which corroborates the work of Araujo and Siqueira (2015).

Additional results also suggest that after the implementation of EC 86, which aimed to give greater responsibility to managers, there was no reduction in the volume of Capital and Current Expenses of the municipalities. In this case, hypothesis H5 is not satisfied, which suggests that EC86 may not have fully contributed to the improvement of the parliamentary amendments process in the budget. In addition, other analyzes suggest that the targeting of voluntary transfers from the Union may be taking place impartially as they continue to be related to the alignment between political parties. This result leads us to reflect that those new measures need to be taken in order to effectively control the distribution of regional public resources.

In general, the results presented provide evidence that contributes by empirically demonstrating that the coalition between the president's party and the mayor's party increases the volume of capital expenditures. The results also contribute by bringing evidence not yet recorded in the Brazilian literature on the fact that changes in tax complexity can favor investment in structural goods insofar as both fiscal complexity and fiscal visibility tend to reduce capital expenditures of studied municipalities. The same results are not obtained for current expenses. An additional contribution is by empirically showing that, in Brazil, the implementation of EC 86, which aims to give greater responsibility to managers, did not reduce the volume of Capital and Current Expenses, as can be seen by the estimated non-significant coefficient in equation 2. Together, these results raise concerns as, in general, the results indicate that capital expenditures have increased above the economy's ability to provide resources and associated with partisan issues, in which laws have not been capable of halting such advances.